GitHub - mikeangstadt/clawforce: Dispatch a mob. Collect the proof. Crowdsourced physical task orchestration engine.

 ██████╗██╗      █████╗ ██╗    ██╗███████╗ ██████╗ ██████╗  ██████╗███████╗
██╔════╝██║     ██╔══██╗██║    ██║██╔════╝██╔═══██╗██╔══██╗██╔════╝██╔════╝
██║     ██║     ███████║██║ █╗ ██║█████╗  ██║   ██║██████╔╝██║     █████╗
██║     ██║     ██╔══██║██║███╗██║██╔══╝  ██║   ██║██╔══██╗██║     ██╔══╝
╚██████╗███████╗██║  ██║╚███╔███╔╝██║     ╚██████╔╝██║  ██║╚██████╗███████╗
 ╚═════╝╚══════╝╚═╝  ╚═╝ ╚══╝╚══╝ ╚═╝      ╚═════╝ ╚═╝  ╚═╝ ╚═════╝╚══════╝

Dispatch a mob. Collect the proof.

ClawForce is a task orchestration engine that gives MCP-compatible AI agents autonomous human deployments for anything in the physical world. Think Terraform, but instead of provisioning servers, you're provisioning people.

Ships with DoorDash, Uber, and TaskRabbit providers out of the box — but any gig platform, errand service, or freelance network is just one interface away.

It gives AI agents like OpenClaw something they've never had: human minions. Your AI reasons about what needs to happen in the physical world. ClawForce makes it happen — from enterprise campaigns to personal errands. One API call.

Pick out a straight 2x4 at Home Depot. Wait in line at Franklin's BBQ so you don't have to. Pick up and drop off your dry cleaning. Photograph 1,000 billboards. Knock 10,000 doors. Inspect 50,000 properties. If a human can do it, ClawForce can dispatch one — or ten thousand.

clawforce campaign create \
  --name "Times Square Flash Mob" \
  --type verification \
  --provider auto \
  --targets locations.csv \
  --instructions "Show up. Dance. Film it."

Why this matters

Every industry that dispatches humans to locations is paying for two things: the person and the orchestration. The person costs $8-80. The orchestration — the scheduling, the management layer, the QA, the reporting — costs 10x more. ClawForce makes the orchestration free. What's left is just the cost of a human showing up.

The other unlock is speed. An insurance company can't inspect 10,000 properties in a day. A retail brand can't audit 5,000 shelves before Monday. An ad agency can't verify 500 billboard placements during a single flight. With ClawForce, all of these are one command, executed in parallel, results aggregated automatically.

Everyday tasks your AI can now handle

Your AI assistant doesn't have hands. ClawForce gives it hands.

Errands & Pickups

# "Hey Claude, I need a straight 2x4 from Home Depot — make sure it's not warped"
clawforce campaign create \
  --name "Lumber Pickup" \
  --type custom \
  --provider taskrabbit \
  --targets '{"address": "Home Depot, 1200 N Lamar Blvd, Austin, TX"}' \
  --instructions "Go to the lumber aisle. Pick out a single 8ft 2x4 — sight down the length to make sure it's straight, no warping, no splits. Purchase it. Deliver to drop-off address."

Wait in Line

# "Get me Franklin's BBQ — I'm not waiting 3 hours"
clawforce campaign create \
  --name "Franklin's BBQ Pickup" \
  --type custom \
  --provider taskrabbit \
  --targets '{"address": "Franklin Barbecue, 900 E 11th St, Austin, TX 78702"}' \
  --instructions "Get in line. Order 1lb brisket, 1lb pulled pork, 1 pint of slaw. Pay with the card on file. Deliver to drop-off address. Keep it warm."

Dry Cleaning, Returns, Drop-offs

clawforce campaign create \
  --name "Dry Cleaning Roundtrip" \
  --type custom \
  --provider doordash \
  --targets '{"address": "456 Elm St, Austin, TX"}' \
  --instructions "Pick up 3 shirts and 1 suit from 456 Elm St (hanging on front door hook). Drop off at EcoClean Dry Cleaners, 789 Congress Ave. Request 2-day turnaround under name 'Angstadt'. Photograph the claim ticket."

If a human can do it, your AI can dispatch one. No custom integrations. No new apps. Just a tool call.

Use cases that will make your CFO uncomfortable

Out-of-Home Ad Verification

What it costs now: Verification companies charge $50-150/placement/month. Manual audits run $200-500/location. What ClawForce costs: $7.75/photo via DoorDash. A 1,000-billboard audit drops from $200K to $8K.

clawforce campaign create \
  --name "Q2 Billboard Audit - Interstate Corridor" \
  --type photo_capture \
  --provider doordash \
  --targets billboards.csv \
  --instructions "Photograph the billboard face. Include surrounding street context. Capture any damage or obstruction."

Insurance Property Inspections

What it costs now: $75-350 per drive-by inspection. Carriers spend billions annually on field inspections. What ClawForce costs: $8-15/property. A 10,000-property book review drops from $2M to $100K.

clawforce campaign create \
  --name "FL Hurricane Portfolio Review" \
  --type photo_capture \
  --provider auto \
  --targets properties.csv \
  --instructions "Photograph front, left side, right side, and roof from street. Note visible damage, missing shingles, standing water, or debris."

Retail Shelf Audits & Planogram Compliance

What it costs now: $30-100/store visit through companies like Trax or Repsly. CPG brands spend $50-100K/month. What ClawForce costs: $8-20/store. A 5,000-store national audit: $40K vs $300K.

clawforce campaign create \
  --name "Walmart Endcap Verification - New Product Launch" \
  --type photo_capture \
  --provider auto \
  --targets walmart-locations.csv \
  --instructions "Go to aisle 7 endcap. Photograph the full display. Capture price tags. Note if product is present, out of stock, or incorrectly placed."

Proof-of-Play for Restaurant/Bar Ad Campaigns

What it costs now: No scalable solution exists. Brands trust the network and hope. What ClawForce costs: $8-15/location. 515 Texas Roadhouse and Buffalo Wild Wings locations verified during a 1-hour ad flight for $4K-$8K.

clawforce compare \
  --type photo_capture \
  --targets restaurant-locations.csv \
  --window 60

# Provider          | Per Task         | Total              | Status
# doordash          | $7.75 - $15.00   | $3,991 - $7,725    | Ready
# taskrabbit        | $20.00 - $80.00  | $10,300 - $41,200  | Stub
# field-nation      | $50.00 - $200.00 | $25,750 - $103,000 | Stub

Political Canvassing & Voter Outreach

What it costs now: Professional canvassers cost $15-25/hour plus management overhead. A statewide petition campaign runs $1-5 per signature. What ClawForce costs: $8-15/door. 10,000 doors knocked in a single Saturday. No hiring, no training, no payroll.

clawforce campaign create \
  --name "Voter Registration Drive - Swing Districts" \
  --type delivery \
  --provider auto \
  --targets voter-addresses.csv \
  --template registration-packet.json \
  --instructions "Deliver voter registration packet. Ring doorbell. If no answer, leave at door and photograph placement."

Legal Process Serving

What it costs now: $50-150 per serve attempt. Specialized servers charge $200+. What ClawForce costs: $8-15/attempt. A firm serving 500 defendants drops from $50K to $6K.

Real Estate Portfolio Inspections

What it costs now: BPOs (Broker Price Opinions) run $50-150 each. Banks order millions annually. What ClawForce costs: $8-15/property. A bank reviewing 50,000 properties in a distressed portfolio: $400K vs $5M.

Mystery Shopping

What it costs now: $25-75 per shop visit. National programs run $500K+/year. What ClawForce costs: $20-35/visit via TaskRabbit. A 2,000-location quarterly mystery shop: $40K vs $150K.

clawforce campaign create \
  --name "Q2 Mystery Shop - Drive-Thru Speed" \
  --type custom \
  --provider taskrabbit \
  --targets franchise-locations.csv \
  --instructions "Visit drive-thru. Order a #3 combo. Time from order to receipt. Photograph the receipt showing timestamp. Rate friendliness 1-5. Upload receipt photo and notes."

Guerrilla Marketing / Street Teams

What it costs now: Agencies charge $25-50/hour per brand ambassador, 4-hour minimums. A 100-city activation runs $500K+. What ClawForce costs: $15-30/person. Fan out 500 street team members in one API call.

Environmental & Compliance Monitoring

What it costs now: Environmental consulting firms charge $150-500/site visit. What ClawForce costs: $8-20/site. A developer with 200 active construction sites saves $60K/quarter on stormwater compliance photos alone.

Disaster Response Documentation

What it costs now: Insurance adjusters fly in at $500-1,000/day. FEMA assessments take weeks. What ClawForce does: Immediately after a disaster, fan out to every affected address. Thousands of damage assessments in hours instead of weeks. The speed alone is worth more than the cost savings.

# Hurricane makes landfall at 2am. By 8am:
clawforce campaign create \
  --name "Hurricane Milton - Immediate Damage Assessment" \
  --type photo_capture \
  --provider auto \
  --targets insured-properties-fl.csv \
  --concurrency 50 \
  --instructions "Photograph property from all accessible angles. Document roof damage, flooding, debris, broken windows. Include street-level context. Do NOT enter damaged structures."

The math

Industry	Traditional Cost	ClawForce Cost	Savings
Billboard audit (1,000 locations)	$200,000	$8,000	96%
Property inspections (10,000)	$2,000,000	$100,000	95%
Retail shelf audit (5,000 stores)	$300,000	$40,000	87%
Proof-of-play verification (515)	N/A (didn't exist)	$4,000	New capability
Process serving (500 defendants)	$50,000	$6,000	88%
Real estate BPOs (50,000)	$5,000,000	$400,000	92%
Mystery shopping (2,000/quarter)	$150,000	$40,000	73%
Disaster assessment (5,000 properties)	Weeks + $500K	Hours + $50K	90% + 100x faster

Install

Prerequisites

Node.js 22+ — install
pnpm — npm install -g pnpm (or corepack enable && corepack prepare pnpm@latest --activate)

That's it. No databases to install, no Redis, no Docker required. ClawForce uses SQLite — the database is a single file created automatically on first run.

Quick start

# Clone
git clone https://github.com/mikeangstadt/clawforce.git
cd clawforce

# Install
pnpm install

# Configure
cp .env.example .env
# Edit .env with your provider credentials (see below)

# Verify it works
pnpm dev --help
pnpm dev providers

# Run a test campaign with the mock provider (no credentials needed)
pnpm dev campaign create \
  --name "Test Campaign" \
  --type photo_capture \
  --provider mock \
  --targets test/fixtures/sample-targets.csv \
  --instructions "Take a photo of the location"

# Check status
pnpm dev campaign list

Configuration

All settings via environment variables or .env file:

# Core (all optional — sensible defaults)
CLAWFORCE_DB_PATH=./clawforce.db         # SQLite database location
CLAWFORCE_PORT=3100                      # HTTP/REST server port
CLAWFORCE_POLL_INTERVAL_MS=30000         # Status poll frequency (ms)
CLAWFORCE_DEFAULT_CONCURRENCY=5          # Max parallel dispatches
CLAWFORCE_LOG_LEVEL=info                 # debug, info, warn, error

# DoorDash Drive API (get from https://developer.doordash.com)
DOORDASH_DEVELOPER_ID=your-developer-id
DOORDASH_KEY_ID=your-key-id
DOORDASH_SIGNING_SECRET=your-signing-secret

# TaskRabbit / Dolly (request access at developer.taskrabbit.com)
TASKRABBIT_AUTH0_DOMAIN=your-dolly-tenant.us.auth0.com
TASKRABBIT_CLIENT_ID=your-client-id
TASKRABBIT_CLIENT_SECRET=your-client-secret
TASKRABBIT_AUDIENCE=your-audience
TASKRABBIT_CLIENT_ENTITY_ID=clawforce
TASKRABBIT_STORE_ID=clawforce-default
TASKRABBIT_BASE_URL=https://papi.sandbox.dolly.com

# Uber Direct (sign up at https://direct.uber.com)
UBER_DIRECT_CUSTOMER_ID=your-customer-id
UBER_DIRECT_CLIENT_ID=your-client-id
UBER_DIRECT_CLIENT_SECRET=your-client-secret

# Field Nation (requires contract — developer.fieldnation.com)
FIELD_NATION_CLIENT_ID=your-client-id
FIELD_NATION_CLIENT_SECRET=your-client-secret
FIELD_NATION_USERNAME=your-username
FIELD_NATION_PASSWORD=your-password
FIELD_NATION_BASE_URL=https://api-sandbox.fndev.net

No credentials? No problem. The mock provider works with zero configuration — use it for development, testing, and demos.

DoorDash sandbox: Sign up at developer.doordash.com, create API keys, and test with their sandbox environment for free before going live.

Running ClawForce

There are four ways to run it, depending on your use case:

1. CLI (direct commands)

# During development (no build step)
pnpm dev <command>

# After building
pnpm build
node dist/index.js <command>

2. Global CLI install

pnpm build
pnpm link --global

# Now available everywhere
clawforce providers
clawforce campaign create ...
clawforce compare --type photo_capture --targets locations.csv --window 60

3. MCP server (for Claude Code, OpenClaw, etc.)

# Run standalone
pnpm mcp

Or add to your MCP client config:

{
  "mcpServers": {
    "clawforce": {
      "command": "npx",
      "args": ["tsx", "/path/to/clawforce/src/mcp/server.ts"],
      "env": {
        "CLAWFORCE_DB_PATH": "/path/to/clawforce.db"
      }
    }
  }
}

Now your AI agent has human minions. It can reason about what needs to happen in the physical world, compare providers and costs, dispatch thousands of people, and collect the results — all through tool calls. The AI plans. ClawForce executes. Humans deliver.

4. HTTP server (REST API)

pnpm server
# ClawForce server running on http://localhost:3100
# REST API: http://localhost:3100/api/
# Health check: http://localhost:3100/health

Includes the status poller — active campaigns are automatically monitored for completion.

Run the tests

pnpm test           # Run once
pnpm test:watch     # Watch mode

How it works

You (or your AI agent) define a campaign
        |
        v
  Template + Targets (CSV/JSON)
        |
        v
  +-----------+
  |  FAN OUT  |  p-queue, concurrency-controlled, per-task error isolation
  +-----------+
        |
   +---------+---------+---------+
   |         |         |         |
   v         v         v         v
 Task 1    Task 2    Task 3    Task N
   |         |         |         |
   v         v         v         v
 Provider  Provider  Provider  Provider
(DoorDash) (TaskRabbit) (Uber) (any)
   |         |         |         |
   v         v         v         v
 Human     Human     Human     Human
 Agent     Agent     Agent     Agent
   |         |         |         |
   v         v         v         v
  +---------------------------+
  |       RE-AGGREGATE        |  poll, collect, milestone webhooks
  +---------------------------+
        |
        v
  Unified results: photos, costs, success rates, provider breakdown

Providers

ClawForce doesn't care who does the work. Every gig platform is just a provider implementing one interface. The engine routes tasks to the cheapest capable provider automatically.

Provider	Task Types	Errand Categories	Coverage	Cost/Task	Status
mock	everything	all	everywhere	$1-5	Ready (dev/test)
doordash	delivery, photo, errand	pickup_dropoff, food_delivery	US (excl. CA, NYC, SEA, CO)	$7.75-15	Ready (Drive API)
taskrabbit	delivery, errand	pickup_dropoff, food_delivery	US, UK, CA, FR, DE, ES	$20-80	Ready (Dolly API)
uber-direct	delivery, errand	pickup_dropoff, food_delivery	US, CA, MX, BR, AU, JP, GB, FR, DE	$5-12	Ready (Direct API)
field-nation	verification, survey, photo, errand, custom	inspection, skilled_labor, multi_step	US	$50-200	Ready (REST v2 API)

*DoorDash photo capture works by dispatching a delivery with specific dropoff_instructions and collecting the verification photo. Creative? Yes. Does it work? Also yes.

Errand category routing

When you dispatch an errand with --provider auto, ClawForce routes based on the errand category — not just the task type. Need someone to pick out a straight 2x4? That's shopping → TaskRabbit. Need a food pickup delivered? That's food_delivery → DoorDash (cheapest). Need furniture assembled? That's skilled_labor → TaskRabbit or Field Nation.

Category	What it means	Best provider
`shopping`	Buy something — requires judgment (quality, selection)	TaskRabbit
`wait_in_line`	Wait at a location, purchase/collect, deliver	TaskRabbit
`pickup_dropoff`	Pick up from A, deliver to B	DoorDash (cheapest), Uber, TaskRabbit
`food_delivery`	Order/pick up food and deliver	DoorDash, Uber
`inspection`	Go look at something and report back	TaskRabbit, Field Nation
`personal_errand`	General personal tasks	TaskRabbit
`multi_step`	2+ stops or sequential steps	TaskRabbit, Field Nation
`skilled_labor`	Assembly, installation, repair	TaskRabbit, Field Nation

# Auto-routes to TaskRabbit (shopping requires judgment + worker_rating)
clawforce campaign create \
  --name "Lumber Run" \
  --type errand \
  --provider auto \
  --errand-category shopping \
  --purchase-budget 2000 \
  --requires-judgment \
  --targets '{"address": "Home Depot, Austin TX"}' \
  --instructions "Buy one 8ft 2x4. Sight down the length — no warping, no splits."

# Auto-routes to DoorDash (food_delivery, cheapest)
clawforce campaign create \
  --name "Lunch Pickup" \
  --type errand \
  --provider auto \
  --errand-category food_delivery \
  --targets '{"address": "Chick-fil-A, 1234 Main St"}' \
  --instructions "Order a #1 combo with lemonade. Deliver to drop-off address."

Compare providers before you commit

clawforce compare --type photo_capture --targets locations.csv --window 60

Cross-provider cost comparison with time window analysis, coverage gap detection, and a recommendation. Available as CLI, MCP tool (compare_estimates), and REST endpoint.

Auto-routing

Set --provider auto and ClawForce picks the best provider per-task based on task type, location, and cost:

clawforce campaign create \
  --name "Nationwide Billboard Audit" \
  --type photo_capture \
  --provider auto \
  --targets billboards.csv \
  --instructions "Photograph the billboard face. Include surrounding context."

NYC target? Routes to TaskRabbit (DoorDash excludes NYC). Rural Texas? Field Nation. Chicago? DoorDash (cheapest with photo support).

Adding a provider

One file. That's it.

// src/providers/my-platform.ts
export class MyPlatformProvider implements TaskProvider {
  name = 'my-platform';
  capabilities = {
    taskTypes: ['delivery', 'photo_capture'],
    features: ['real_time_tracking', 'custom_instructions'],
    coverage: { countries: ['US'] },
    maxConcurrency: 10,
    estimatedCostRange: { minCents: 500, maxCents: 1500 },
  };

  async dispatch(task, template) { /* hit your API */ }
  async getStatus(id) { /* check status */ }
  async cancel(id) { /* cancel it */ }
  extractResult(data) { /* pull photos, fees, etc. */ }
  validateTemplate(t) { /* check required fields */ }
}

Register it in src/providers/registry.ts. Done. The engine, CLI, MCP tools, and REST API all pick it up automatically.

CLI Reference

List providers

# All providers
clawforce providers

# Only providers that support photo capture
clawforce providers --type photo_capture

Compare costs across providers

# Compare all providers for a task type
clawforce compare --type photo_capture --targets locations.csv

# With a time window constraint (e.g., 1-hour ad flight)
clawforce compare --type photo_capture --targets locations.csv --window 60

Create a campaign

clawforce campaign create \
  --name "Postcard Drop Q2" \
  --type delivery \
  --provider doordash \
  --targets addresses.csv \
  --template template.json \
  --concurrency 10 \
  --delay 200 \
  --webhook https://your-server.com/hooks/clawforce

# Dry run (creates records, doesn't dispatch)
clawforce campaign create \
  --name "Test Run" \
  --type delivery \
  --provider mock \
  --targets addresses.csv \
  --dry-run

Track progress

clawforce campaign status <campaign_id>
clawforce campaign list
clawforce campaign list --status active

Get results

# Summary
clawforce campaign results <campaign_id>

# Per-task details
clawforce campaign results <campaign_id> --details

# CSV export
clawforce campaign results <campaign_id> --details --format csv > results.csv

Estimate cost (single provider)

clawforce estimate \
  --type delivery \
  --provider doordash \
  --targets addresses.csv \
  --template template.json

Cancel

clawforce campaign cancel <campaign_id>

Inspect a single task

clawforce task <task_id>

MCP Tools

When running as an MCP server, ClawForce exposes these tools:

Tool	What it does
`list_providers`	Show available providers, capabilities, coverage, costs
`compare_estimates`	Compare costs across ALL providers for a task type with time window analysis
`create_campaign`	Create + fan out a task campaign
`get_campaign_status`	Progress metrics, provider breakdown
`get_results`	Aggregated results: photos, costs, per-task details
`cancel_campaign`	Kill it
`list_campaigns`	List campaigns by status
`estimate_campaign`	Cost estimate from a single provider

Your AI agent can now say "compare the cost of sending someone to photograph all 515 of these restaurant locations during a 1-hour window, then use the cheapest provider that covers all locations" — and ClawForce will do exactly that.

REST API

Start the server:

pnpm server
# ClawForce server running on http://localhost:3100

Method	Endpoint	Maps to
GET	`/api/providers`	`list_providers`
POST	`/api/compare`	`compare_estimates`
POST	`/api/campaigns`	`create_campaign`
GET	`/api/campaigns`	`list_campaigns`
GET	`/api/campaigns/:id/status`	`get_campaign_status`
GET	`/api/campaigns/:id/results`	`get_results`
POST	`/api/campaigns/:id/cancel`	`cancel_campaign`
POST	`/api/estimate`	`estimate_campaign`
GET	`/health`	Health check

Target files

CSV

address,name,phone
"123 Main St, Springfield, IL 62701",John Smith,555-0101
"456 Oak Ave, Chicago, IL 60601",Jane Doe,555-0102

JSON

[
  { "address": "123 Main St, Springfield, IL 62701", "name": "John Smith", "phone": "555-0101" },
  { "address": "456 Oak Ave, Chicago, IL 60601", "name": "Jane Doe", "phone": "555-0102" }
]

Extra columns in CSV become metadata on the target — use them for campaign-specific data.

Template files

{
  "pickupAddress": "100 Print Shop Lane, Springfield, IL 62701",
  "pickupBusinessName": "FastPrint Co",
  "pickupPhoneNumber": "555-0000",
  "pickupInstructions": "Postcards are in the labeled bin by the front door",
  "dropoffPhoneNumber": "555-9999",
  "dropoffInstructions": "Place in mailbox. If no mailbox, leave at front door.",
  "customInstructions": "Handle with care — these are personalized postcards",
  "orderValue": 100,
  "tip": 200
}

Configuration

All via environment variables (or .env file):

Variable	Default	What it does
`CLAWFORCE_DB_PATH`	`./clawforce.db`	SQLite database location
`CLAWFORCE_PORT`	`3100`	HTTP server port
`CLAWFORCE_POLL_INTERVAL_MS`	`30000`	How often to poll provider status (ms)
`CLAWFORCE_DEFAULT_CONCURRENCY`	`5`	Default parallel dispatch limit
`CLAWFORCE_LOG_LEVEL`	`info`	Log level (debug, info, warn, error)
`DOORDASH_DEVELOPER_ID`	--	DoorDash Drive API credentials
`DOORDASH_KEY_ID`	--
`DOORDASH_SIGNING_SECRET`	--
`UBER_DIRECT_CUSTOMER_ID`	--	Uber Direct API credentials
`UBER_DIRECT_CLIENT_ID`	--
`UBER_DIRECT_CLIENT_SECRET`	--
`TASKRABBIT_AUTH0_DOMAIN`	--	TaskRabbit/Dolly API credentials
`TASKRABBIT_CLIENT_ID`	--
`TASKRABBIT_CLIENT_SECRET`	--
`TASKRABBIT_AUDIENCE`	--
`TASKRABBIT_BASE_URL`	`https://papi.sandbox.dolly.com`	Sandbox/production URL
`FIELD_NATION_CLIENT_ID`	--	Field Nation API credentials
`FIELD_NATION_CLIENT_SECRET`	--
`FIELD_NATION_USERNAME`	--	Field Nation login username
`FIELD_NATION_PASSWORD`	--	Field Nation login password
`FIELD_NATION_BASE_URL`	`https://api-sandbox.fndev.net`	Sandbox/production URL

Architecture

src/
  providers/        # Provider interface + implementations
    interface.ts    # The contract: capabilities, dispatch, getStatus, cancel, extractResult
    registry.ts     # Capability-based routing + auto-resolution + compare
    mock.ts         # Dev/test (simulates full lifecycle)
    doordash.ts     # DoorDash Drive API
    taskrabbit.ts   # TaskRabbit/Dolly Delivery API
    uber-direct.ts  # Uber Direct API
    field-nation.ts # Field Nation REST v2 API
  engine/           # Core orchestration
    fanout.ts       # Template x Targets -> dispatched tasks (p-queue)
    poller.ts       # Status polling loop
    aggregator.ts   # Result collection + milestone webhooks
  mcp/              # MCP server (8 tools)
  cli/              # CLI commands
  db/               # SQLite + Drizzle schema
  models/           # Campaign, Task, Result CRUD

Single process. SQLite. No Redis. No Kafka. No Kubernetes. Just a binary that dispatches humans.

License

MIT

Built with Claude Code

Name		Name	Last commit message	Last commit date
Latest commit History 22 Commits
src		src
test		test
.env.example		.env.example
.gitignore		.gitignore
.pnpmrc.json		.pnpmrc.json
README.md		README.md
clawforce.db-shm		clawforce.db-shm
drizzle.config.ts		drizzle.config.ts
package.json		package.json
pnpm-lock.yaml		pnpm-lock.yaml
tsconfig.json		tsconfig.json

Folders and files

Latest commit

History

Repository files navigation

Why this matters

Everyday tasks your AI can now handle

Errands & Pickups

Wait in Line

Dry Cleaning, Returns, Drop-offs

Use cases that will make your CFO uncomfortable

Out-of-Home Ad Verification

Insurance Property Inspections

Retail Shelf Audits & Planogram Compliance

Proof-of-Play for Restaurant/Bar Ad Campaigns

Political Canvassing & Voter Outreach

Legal Process Serving

Real Estate Portfolio Inspections

Mystery Shopping

Guerrilla Marketing / Street Teams

Environmental & Compliance Monitoring

Disaster Response Documentation

The math

Install

Prerequisites

Quick start

Configuration

Running ClawForce

1. CLI (direct commands)

2. Global CLI install

3. MCP server (for Claude Code, OpenClaw, etc.)

4. HTTP server (REST API)

Run the tests

How it works

Providers

Errand category routing

Compare providers before you commit

Auto-routing

Adding a provider

CLI Reference

List providers

Compare costs across providers

Create a campaign

Track progress

Get results

Estimate cost (single provider)

Cancel

Inspect a single task

MCP Tools

REST API

Target files

CSV

JSON

Template files

Configuration

Architecture

License

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages